natas_hw
/
jidt
mirror of https://github.com/jlizier/jidt
0
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Pulled some common functionality for testing multivariate TE calculators into an abstract class. 

Added basic tester for Kraskov multivariate TE
This commit is contained in:
joseph.lizier 2012-12-13 14:50:33 +00:00
parent e789acd4dc
commit a6e09c7fc4
3 changed files with 140 additions and 58 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

86
java/unittests/infodynamics/measures/continuous/TransferEntropyMultiVariateTester.java Executable file
View File

@ -0,0 +1,86 @@
package infodynamics.measures.continuous;
import junit.framework.TestCase;
import infodynamics.utils.MatrixUtils;
import infodynamics.utils.RandomGenerator;
public abstract class TransferEntropyMultiVariateTester extends TestCase {
/**
* Confirm that the local values average correctly back to the average value
*
* @param teCalc a pre-constructed TransferEntropyCalculatorMultiVariate object
* @param dimensions number of dimensions for the source and dest data to use
* @param timeSteps number of time steps for the random data
* @param k history length for the TE calculator to use
*/
public void testLocalsAverageCorrectly(TransferEntropyCalculatorMultiVariate teCalc,
int dimensions, int timeSteps, int k)
throws Exception {
teCalc.initialise(k, dimensions, dimensions);
// generate some random data
RandomGenerator rg = new RandomGenerator();
double[][] sourceData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
double[][] destData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
teCalc.setObservations(sourceData, destData);
//teCalc.setDebug(true);
double te = teCalc.computeAverageLocalOfObservations();
//teCalc.setDebug(false);
double[] teLocal = teCalc.computeLocalOfPreviousObservations();
System.out.printf("Average was %.5f\n", te);
assertEquals(te, MatrixUtils.mean(teLocal, k, timeSteps-k), 0.00001);
}
/**
* Confirm that significance testing doesn't alter the average that
* would be returned.
*
* @param teCalc a pre-constructed TransferEntropyCalculatorMultiVariate object
* @param dimensions number of dimensions for the source and dest data to use
* @param timeSteps number of time steps for the random data
* @param k history length for the TE calculator to use
* @throws Exception
*/
public void testComputeSignificanceDoesntAlterAverage(TransferEntropyCalculatorMultiVariate teCalc,
int dimensions, int timeSteps, int k) throws Exception {
teCalc.initialise(k, dimensions, dimensions);
// generate some random data
RandomGenerator rg = new RandomGenerator();
double[][] sourceData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
double[][] destData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
teCalc.setObservations(sourceData, destData);
//teCalc.setDebug(true);
double te = teCalc.computeAverageLocalOfObservations();
//teCalc.setDebug(false);
//double[] teLocal = teCalc.computeLocalOfPreviousObservations();
System.out.printf("Average was %.5f\n", te);
// Now look at statistical significance tests
int[][] newOrderings = rg.generateDistinctRandomPerturbations(
timeSteps - k, 100);
teCalc.computeSignificance(newOrderings);
// And compute the average value again to check that it's consistent:
for (int i = 0; i < 10; i++) {
double averageCheck1 = teCalc.computeAverageLocalOfObservations();
assertEquals(te, averageCheck1);
}
}
}

62
java/unittests/infodynamics/measures/continuous/kernel/TransferEntropyMultiVariateTester.java
View File

@ -1,12 +1,8 @@
package infodynamics.measures.continuous.kernel;
import infodynamics.utils.MatrixUtils;
import infodynamics.utils.RandomGenerator;
import junit.framework.TestCase;
public class TransferEntropyMultiVariateTester
extends infodynamics.measures.continuous.TransferEntropyMultiVariateTester {
public class TransferEntropyMultiVariateTester extends TestCase {
/**
* Confirm that the local values average correctly back to the average value
*
@ -16,9 +12,6 @@ public class TransferEntropyMultiVariateTester extends TestCase {
TransferEntropyCalculatorMultiVariateKernel teCalc =
new TransferEntropyCalculatorMultiVariateKernel();
int dimensions = 2;
int timeSteps = 100;
int k = 1;
String kernelWidth = "1";
teCalc.setProperty(
@ -27,25 +20,8 @@ public class TransferEntropyMultiVariateTester extends TestCase {
teCalc.setProperty(
TransferEntropyCalculatorMultiVariateKernel.EPSILON_PROP_NAME,
kernelWidth);
teCalc.initialise(k, dimensions, dimensions);
// generate some random data
RandomGenerator rg = new RandomGenerator();
double[][] sourceData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
double[][] destData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
teCalc.setObservations(sourceData, destData);
//teCalc.setDebug(true);
double te = teCalc.computeAverageLocalOfObservations();
//teCalc.setDebug(false);
double[] teLocal = teCalc.computeLocalOfPreviousObservations();
System.out.printf("Average was %.5f\n", te);
assertEquals(te, MatrixUtils.mean(teLocal, k, timeSteps-k), 0.0001);
super.testLocalsAverageCorrectly(teCalc, 2, 100, 1);
}
/**
@ -59,9 +35,6 @@ public class TransferEntropyMultiVariateTester extends TestCase {
TransferEntropyCalculatorMultiVariateKernel teCalc =
new TransferEntropyCalculatorMultiVariateKernel();
int dimensions = 2;
int timeSteps = 100;
int k = 1;
String kernelWidth = "1";
teCalc.setProperty(
@ -70,35 +43,8 @@ public class TransferEntropyMultiVariateTester extends TestCase {
teCalc.setProperty(
TransferEntropyCalculatorMultiVariateKernel.EPSILON_PROP_NAME,
kernelWidth);
teCalc.initialise(k, dimensions, dimensions);
// generate some random data
RandomGenerator rg = new RandomGenerator();
double[][] sourceData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
double[][] destData = rg.generateNormalData(timeSteps, dimensions,
0, 1);
teCalc.setObservations(sourceData, destData);
//teCalc.setDebug(true);
double te = teCalc.computeAverageLocalOfObservations();
//teCalc.setDebug(false);
//double[] teLocal = teCalc.computeLocalOfPreviousObservations();
System.out.printf("Average was %.5f\n", te);
// Now look at statistical significance tests
int[][] newOrderings = rg.generateDistinctRandomPerturbations(
timeSteps - k, 100);
teCalc.computeSignificance(newOrderings);
// And compute the average value again to check that it's consistent:
for (int i = 0; i < 10; i++) {
double averageCheck1 = teCalc.computeAverageLocalOfObservations();
assertEquals(te, averageCheck1);
}
super.testComputeSignificanceDoesntAlterAverage(teCalc, 2, 100, 1);
}
}

50
java/unittests/infodynamics/measures/continuous/kraskov/TransferEntropyMultiVariateTester.java Executable file
View File

@ -0,0 +1,50 @@
package infodynamics.measures.continuous.kraskov;
public class TransferEntropyMultiVariateTester
extends infodynamics.measures.continuous.TransferEntropyMultiVariateTester {
/**
* Confirm that the local values average correctly back to the average value
*
*/
public void testLocalsAverageCorrectly() throws Exception {
TransferEntropyCalculatorMultiVariateKraskov teCalc =
new TransferEntropyCalculatorMultiVariateKraskov();
String kraskov_K = "4";
teCalc.setProperty(
TransferEntropyCalculatorMultiVariateKraskov.PROP_KRASKOV_ALG_NUM,
"2");
teCalc.setProperty(
MutualInfoCalculatorMultiVariateKraskov.PROP_K,
kraskov_K);
super.testLocalsAverageCorrectly(teCalc, 2, 100, 1);
}
/**
* Confirm that significance testing doesn't alter the average that
* would be returned.
*
* @throws Exception
*/
public void testComputeSignificanceDoesntAlterAverage() throws Exception {
TransferEntropyCalculatorMultiVariateKraskov teCalc =
new TransferEntropyCalculatorMultiVariateKraskov();
String kraskov_K = "4";
teCalc.setProperty(
TransferEntropyCalculatorMultiVariateKraskov.PROP_KRASKOV_ALG_NUM,
"2");
teCalc.setProperty(
MutualInfoCalculatorMultiVariateKraskov.PROP_K,
kraskov_K);
super.testComputeSignificanceDoesntAlterAverage(teCalc, 2, 100, 1);
}
}